Network properties of cancer prognostic genes and their gene sets in the human protein interactome
Abstract Background Identifying prognostic genes (PG) is crucial for estimating survival time and providing pinpoint treatments for patients with cancer. However, prognostic genes sets (PGS) reported in most existing research have low reproducibility and overlap ever between the same cancers or their subtypes. Their common characteristic as well as the molecular mechanism of action is still elusive. Methods Here, we obtained nine prognostic gene sets (including 1,439 prognostic genes) of different types of cancer from 23 high quality literatures, and systemically investigated eight network topological properties for PG and PGS compared with background and four other gene sets (cancer gene set CA, essential gene set ES, housekeeping gene set HK, and metastasis-angiogenesis gene set MA) based on the HPRD and String networks. Results The results showed that PG did not occupy key positions in the human protein interactome network, and were more similar to ES rather than CA. Also, PGS had significantly small intraset distance (IAD) and interset distance (IED) in comparison with random sets. Further, we also found that PGS tended to have be distributed within network modules rather than between modules, the functional intersection of the modules enriched with PGS was closely related to cancer. Conclusions Our research reveals the common properties of cancer PG and PGS in the human protein interactome network, and can help us understand and discover cancer prognostic biomarkers.